Trying to deploy in a serverless fashion this framework. The serverless function works locally but still needs some tweaks to work in the cloud. The model folder contains the pretrained model for the function to work. Pull request highly appreciated.
The package is to heavy for aws lambda even with some tweaks by reducing dependencies size (get rid of shared libraries etc). I've not tested on kubeless. I keep the repository up for the records
- Have node and npm installed. There is a good guide for installing node and npm on linux here
- Have python and pip installed. There is a good guide for installing python with conda here
- Have virtualenv installed:
pip install virtualenv
- Make sure you have exported your AWS keys in your environmnet variables
npm install -g serverless
git clone https://github.com/Sach97/serverless-multilabel-text-classification.git
cd serverless-multilabel-text-classification
virtualenv venv --python=python3
source venv/bin/activate
pip3 install git+https://github.com/inspirehep/[email protected] && pip3 install tensorflow
serverless invoke local -f predict --data '{"text":"Stephen Hawking studies black holes"}' --log # uncomment the two lines in handler.py
git clone https://github.com/Sach97/serverless-multilabel-text-classification.git
cd serverless-multilabel-text-classification
chmod +x build_vendored.sh
chmod +x clean_venv.sh
chmod +x env_var.sh
virtualenv venv --python=python3
source venv/bin/activate
pip3 install git+https://github.com/inspirehep/[email protected] && pip3 install tensorflow
sh clean_venv.sh
sh build_vendored.sh
python train_model.py
python upload_model.py
serverless invoke local -f predict --data '{"text":"Stephen Hawking studies black holes"}' --log
serverless invoke local -f predict --data '{"text":"Stephen Hawking studies black holes"}' --log
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent